50
This file was created by scanning the printed publication.
Errors identified by the software have been corrected;
however, some errors may remain.
Effects of Fire Behavior on Prescribed Fire Smoke Characteristics:
A Case Study
Wayne Einfeld, Darold E. Ward, and Colin Hardy
Biomass burning on a global scale injects a substantial
quantity of gaseous and particulate matter emissions
into the troposphere. Some of these combustion
products are known to accumulate in the atmosphere
and are implicated in observed changes in tropospheric composition and chemistry. The practice of
open burning of biomass has come under close examination as a result of its pollution potential. In the
United States, most biomass burning is managed in
order to limit pollutant release. Recently, it has been
inferred that biomass burning may produce a net
increase in the concentration of "greenhouse gases"
and particles in the atmosphere. Estimates of the
quantity of biomass consumed globally per year vary
widely, but generally Seiler and Crutzen's (1980) estimate of 5 x 1015 g yr- 1 (grams per year) is regarded as
accurate to within a factor of 2. The degree of consumption of total available biomass is affected by the
quantity of biomass present, size distribution of the
fuel, precipitation history for the site, condition of
the vegetation, local meteorology, and a host of other .
factors. Combustion characteristics vary widely from
vegetation types ranging from tropical grasslands to
subarctic forest. As a result, the total emissions released are a complex function of many variables
(Ward, 1990). Factors contributing to the total release
of a particular pollutant species from a biomass fire
can be quantitatively described by the following
expression:
Mx
~
Mt
X
A
X
[c
X
EFx
where Mx is the mass yield of a particular gaseous or
particulate species, M1 is the fuel load in mass per unit
area, A is the area burned, f, is the fractional consumption of the total fuel load, and EFx is the emission factor for the species of interest. Here, emission
factor is defined as the mass release of a particular
species per unit mass of fuel consumed. Ward and
Hardy (1984) showed that emission factors for particulate matter and gaseous products are functions of the
phase of combustion (flaming or smoldering). The
work described here further demonstrates the impor-
tance of combustion phase on the release of incomplete products of combustion. More specifically, in
this chapter we report results from a study that was
designed to measure all of the terms noted in the
above expression in an effort to derive an estimate of
the total release of important pollutant species from a
well-characterized fire. Ground and aircraft measurements were coordinated to yield a relatively complete
picture of fire behavior and accompanying smoke production. Results from these measurements are then
integrated over the lifetime of the fire and compared
to less rigorous methods of estimating pollutant yield.
Experimental Methods
Fire Site Characteristics and Measurements
A 0.43 square kilometer (km2) site in northwestern
Montana, shown in Figure 50.1, was selected for this
particular study. The fuel consisted of logging slash
resulting from the clear-cut harvesting of true fir and
western larch commercial timber from the site. The
material left on the site was not of commercial value
and provided much of the biomass fuel consumed by
the fire. Prefire site characterization included a fuel
inventory accomplished by taking a random transect
across the site and tallying all fuel material within the
transect into fuel size categories. Extrapolations were
then made to the total area to be burned. Fuel moisture was also determined from a number of wood
sections by specimen weighing, oven drying, and final
weighing. Estimates of fuel consumption were obtained by fastening wires around the circumference of
selected fuel in each size category. Following the fire,
the wires were pulled tight, and the excess wire length
used as a measure of fuel mass consumption in each
fuel category. The fire was lit with hand-carried
torches from the center of the plot outward in concentric circles. An observer was positioned in a suitable
location to map the fire progression on all nine subunits throughout the duration of the fire. Fire characteristics such as the ignition time, duration and extent
413
Einfeld, Ward, and Hardy
Table 50,1
Sampling and analysis methods used on the instrumented aircraft
Analyte or parameter
Analytical method
Particle size and count (0.01-0.5 JJ.ffi)
Grab sample-mobility analyzer
Particle size and count (0.1-3 1-1-rn)
Wing-mounted laser optical spectrometer
Particle size and count (2-32
~-tm)
Wing-mounted laser optical spectrometer
Light scattering coefficient @ 475 nm
Integrating nephelometer
Fine particle ( <3 !LID) mass concentration
Collect particulate on Teflon filter-gravimetric analysis
Fine particle nonvolatile and volatile carbon concentration
Collect particulate on quartz filter--combustion analysis
Carbon dioxide and carbon monoxide
Continuous analyzers--gas filter infrared absorption
DOWN CREEK BURN
1 mile --------~1
LEGEND
~
Area Burned
E:iliili:J (1 07 acres)
T.61 N. R.34W.
SECTION
33
of the flaming phase, and the onset of the smoldering
phase were recorded for each of the nine subunits.
Instrumented Aircraft Measurements
The Sandia National Laboratories instrumented DeHavilland Twin Otter aircraft was used as a sampling
platform during the fire. The aircraft was equipped
with a number of sampling systems as outlined in
Table 50.1. Particle size distributions over the range of
0.01 to 30 micrometer diameter were measured using
mobility and wing-mounted laser optical particle
counters. Gas sampling was carried out for carbon
dioxide and carbon monoxide with continuous monitors using gas filter correlation infrared absorption techniques. Particulate material was collected
through an external probe that extended forward of
the aircraft windshield and terminated in a 1 m3 capacity bag positioned in the rear compartment of the
aircraft, as schematically shown in Figure 50.2. The
bag was filled to capacity in about 5 seconds while the
aircraft was flying through smoke. Immediately following sampling, the grab sample was pumped
through a series of cyclones with a 3 micrometer aerodynamic diameter cut point. The sample was then
routed through a flash lamp nephelometer and finally
through quartz and Teflon filters. Sample volumes
were determined by time integrated measurement of
-
SAMPliNG DUCT
Figure 50.1
subunits.
Map of burn location with inset showing the nine
Figure 50.2 Schematic diagram of the sampling configuration
on the instrumented aircraft.
414
Chapter 50
air flow through each of the filters using mass flow
meters output continuously logged by a data acquisition system. Postsampling analysis on the filters included gravimetric analysis, X-ray fluorescence
elemental analysis, and particulate carbon analysis by
a two-step combustion procedure as given by Johnson
eta!. (1981). Eight smoke samples were collected that
covered all phases of the fire from the initial lighting
phase to the final smoldering phase.
Emission factors for the gases and particulate categories of interest were calculated from samples
collected with the aircraft using a chemical element
balance method (Ward eta!. 1982; Radke eta!., 1988)
whereby total combustion-derived carbon in the
plume is used as a tracer for total mass consumption at
the ground. The concentration ratio of a particular
species to the total combustion-derived carbon concentration in the same parcel of plume air sampled
with the aircraft is corrected to account for the carbon
fraction of the fuel consumed. The resulting expression is given by
X j;
[Ctotad
EF. =[X]
x
where EFx is the emission factor for a given plume
component x, [X] is the measured plume concentration of the component, [ C1o~od is the total combustionderived carbon concentration in the same plume
sample, andf, is the dry fuel carbon fraction, which in
this case is taken to be 0.5 according to Byram (1959).
Table 50.2
Measured fuel loading and consumption
Mass loading
(kg m-')
Fuel diameter (em)
Mass
consumption
0-0.5
0.87
100
0.5-2.5
1.46
100
2.5-7.5
1.25
100
7.5-23
6.81
87
23-50
2.44
50
50+
Total woody fuel
Duff layer
Total woody + duff
0.52
30
13.35
81
10.71
62
24.06
73
DOWN CREEK BURN
"~---
"
16 -
~
14
~
12
~
10
~
~
Composite Are<J in Consumpl!_?~"-------
'''
'
''
8 6 -
20:04
Time
Results
Fuel Loading and Consumption
Results from the pre- and postfire biomass inventories provided a measurement of the fuel consumption
listed in Table 50.2. The total solid wood fuel loading
was estimated at 13.5 kg m- 2 • About70% of the mass
of the large diameter fuel was contained in the 8 to 23
and 23 to 50 em diameter categories. The thick duff
layer consisting of partially decomposed needles,
cones, and twigs covering the forest floor increased
the total fuel loading at the site to 24.1 kgm - 2 • Nearly
all of the fuel was consumed in fuel-size categories
smaller than 23 em. Duff consumption was· estimated
at about 60% of the total duff on the site. Total fuel
consumption (woody fuels, litter, and duff) was 17.6
kg m - 2 , which was 73% of the total biomass fuel on
the site. The percent consumption by fuel categories
and phase of combustion was as follows: 58.5% and
3.4% for the woody fuel for flaming and smoldering
Figure 50.3 A plot of combustion history for each of the nine
subunits. The lighting phase is indicated by a steep linear
increase, the flaming phase by a horizontal segment, and the
smoldering phase by an exponential decay.
phases, respectively; and 12.5% and 25.6% for the
duff fuel for the flaming and smoldering combustion
phases, respectively. The high proportion of total fuel
consumed by this fire was attributed to the low moisture content of the fuels (ranging from 13% to 28%)
at the time of the fire.
Time-Resolved Fire Behavior
Figure 50.3 is a plot of the progress of ignition of the
nine subunits of the plot shown in Figure 50.1. A
ground-based observer periodically drew in isopleth
lines showing the location of the fire front on each of
the subunits. The time of ignition is indicated by the
steep positive-sloped lines in Figure 50.3 for each of
the subunits. The duration of the flaming phase for
415
Einfcld, Ward, and Hardy
70
i
60
60
8;;::
40
~
30
20
10
0
18:00
18:20
18:40
19:00
19:20
19:40
20:00
20:20. 20:40
60
21:00
LOCAL TIME
Figure 50.4
A plot of combustion history with all subunits combined. Dark shading represents flaming combustion and light shading
represents smoldering combustion. Combustion efficiency for each of the eight aircraft smoke samples is shown by black squares.
each subunit is indicated by the flat portion of the
curve, and the smoldering period is shown by the
exponential decay curves for each of the subunits. An
empirically derived half-life of 9.5 minutes (Ward and
Hardy, 1984) is used to define the rate of change for
the fuel consumption during the smoldering combustion phase. A composite merging of the flaming and
smoldering phase of combustion activity from all nine
subunits is shown in Figure 50.4. The period of extensive flaming phase combustion during the first 60
minutes of the fire is of particular significance and is
well correlated witb the characteristics of the emissions produced during that period.
Combustion Efficiency
A term calJed combustion efficiency has been used by
Ward and Hardy (1984) to characterize the completeness of the oxidation of released carbon during the
combustion of biomass fuels. Combustion efficiency
is a particularly sensitive parameter that represents
the degree of contribution from flaming and smoldering combustion in a volume of smoke. It is quantitatively given by the following:
[Ceo,]
Comb eff (%) = -[C] X 100
total
where Cco2 is the background corrected C0 2 carbon
as measured in the plume sample, and C.otal is the sum
of alJ forms of background corrected carbon in the
same plume sample. C,o"' in this case includes carbon
from both gaseous (C02 , CO, methane, and nonmethane hydrocarbons) and particulate species (volatile and nonvolatile carbon) in the plume. The
combustion efficiencies for the eight samples collected throughout the burn are shown in Figure 50.4
superimposed on the time history of the fire to illustrate the changes in combustion efficiency throughout
the fire period. The combustion efficiencies range,
with the exception of one outlier, from about 96% at
the onset of the fire when flaming conditions were·
predominant, to about 85% during smoldering conditions during the later stages of the fire. An observed
83% combustion efficiency for the third sample, although collected early in the fire, suggests that this
particular plume sample originates from smoldering
combustion from a smoke region outside the major
smoke column that was normally sampled with the
aircraft. The range of combustion efficiencies observed during this fire are similar to measurements
made in fires in Canada (Susott et al., 1991), Brazil
(Ward et al., 1991), and the western United States
(Ward and Hardy, 1990).
Particle and Gas Emission Factors
A plot of fine particulate and volatile particulate carbon emission factors is given as a function of sample
combustion efficiency in Figure 50.5. Fine particle
emission factors in the range of 2 to 4 g kg- 1 were
measured during the flaming periods. Increases on
416
Chapter 50
16,------------------------------------,
J..
PM-2.6
•
OC-2.5
200
100
13
160
.'
•
~
0
t"
140-
120
•I
~
z 100
g
w
80.
8
60
~
3
40
20·
00
•
•
00
00
00
•
•
00
00
~
0 +-,-~~·
80
82
•
Cot.'BUSTIOO EFFICIENCY(%)
00
00
00
•
,---,-----,~~~~~-i
•
00
00
-
COM3USTION EFFICIENCY(%)
Figure 50.5 PM2.5 and nonvolatile carbon particulate (OC2.5)
emission factors as a function of combustion efficiency for all
samples collected with the instrumented aircraft.
Figure 50.6 Carbon monoxide emission factors as a function
of combustion efficiency for all samples collected with the
the order of threefold are observed for samples collected during periods when smoldering combustion
was dominant. An opposite although more variable
trend was observed for nonvolatile or elemental carbon measurements where emission factors were high
during flaming combustion and decreased during
smoldering combustion. During flaming combustion,
the fine particles contained about 30% elemental carbon by weight. In the smoldering phase the elemental
carbon content of the fine particles dropped to about
4%. These measurements clearly suggest that
changes in the combustion efficiency of smoke samples are accompanied by changes in the composition
of the particulate matter. During the predominantly
flaming phase, the content of the particulate matter is
low in organic carbon and high in elemental carbon.
During smoldering combustion the opposite relation
appears valid. These observations are consistent with
measurements made by Ward and Hardy (1991) on
the particulate matter content of smoke collected
from both flaming and smoldering combustion using
tower-based sampling systems positioned over large
area fires.
Elements such as potassium, sulfur, iron, and lead
were detected by X-ray fluorescence on most of the
filter samples. Potassium emission factors were relatively constant throughout the duration of the fire
with an average emission factor of 0.107 g kg·'. The
ratio of fine particle mass to fine particle potassium
ranges from about 90 for the flaming periods to 180
for the smoldering periods. Particulate sulfur emissions showed a clear correlation with fine particle
emission factors and ranged from low levels of about
0.03 g kg- 1 during flaming phase to about 0.07 g kg·'
during the smoldering phase. Lead emissions were
relatively uniform over the entire fire period and at a
low level of about 2 mg kg- 1 .
Carbon monoxide emission factors calcnlated for
each of the eight plume samples are plotted as a
function of sample combustion efficiency and are
shown in Figure 50.6. The CO yield can be observed
to increase from about 40 g kg_, to about 120 g kg_,
as the fire progresses into the smoldering phase. A
least squares regression analysis with the CO emission
factor as the dependent variable and combustion efficiency as the independent variable reveals a linear
relationship with a slope of -10.52 (0.58), an intercept of 1043 (7) with a coefficient of determination
(R 2 ) of 0.982.
A summary of differences between smoke emission
factors for selected pollutant species from flaming and
smoldering phase combustion is given in Table 50.3.
An arithmetic mean of the emission factors for the
first two samples was chosen to represent the flaming
phase of the fire. Similarly, the smoldering category is
represented by an average of the final two aircraft
samples collected during the latter stages of the fire.
A comparison of the overall emission factor averages
from all eight aircraft samples with the combustion
phase specific resnlts shows that the overall averages
do not fall at the midpoint between the flaming and
smoldering phases but are closer to the smoldering
valnes. Even though an analysis of the fuel consumption shows that approximately 71% of the fuel was
consumed during the flaming phase, the average
emission factors more closely represent the mix of
emissions that would be expected for the smoldering
phase. Since the rate of fnel consumption during
instrumented aircraft.
417
Einfeld, Ward, and Hardy
Table 50.3 Measured smoke emission factors for flaming and smoldering combustion phases and overall averages
Emission factor (g kg- 1)
Smoldering
Flaming
Species or parameter
Fine particulate ( <3 micron)
Elemental carbon (<3 micron)
3.0
11.5
8.10 (3.47)
0.8
0.5
0.55 (0.27)
3.6
Organic carbon (<3 micron)
11.5
35.0
Carbon monoxide
7.61 (3.19)
1576
1733
Carbon dioxide
Overall
1621 (74)
117.2
96.9 (43.7)
Single scatter albedo
0.75
0.97
0.90 (0.10)
Specific scattering
5.8
9.1
8.00 (1.80)
2
Coefficient@ 480 nm (m g-
1
)
Note: The uncertainty (one standard deviation) for the overall average is given in parentheses.
flaming combustion is high, the time duration of the
flaming phase was short relative to the much longer
smoldering phase, during which the remaining 29%
percent of the fuel was consumed. As a result, emission factor averages from the eight samples collected
over the duration of the fire are skewed toward the
smoldering condition.
Particle Size Distributions
Particle size data collected during the first two and
final two aircraft passes were examined in detail in
order to observe effects of fire behavior on particle
size distributions. The first two aircraft samples were
collected during a flaming phase as illustrated by calculated combustion efficiencies of 0.95 and 0.96. The
last two samples were collected during smoldering
conditions as evidenced by combustion efficiencies of
0.88 and 0.87. Number distributions for these four
passes are shown in Figure 50.7. Although significant
differences between the flaming and smoldering samples are not observed, some subtle differences can be
noted. The samples from flaming combustion show an
elevated condensation nuclei mode in the 0.02 to 0.04
micron particle diameter range when compared to the
smoldering phase distributions. An upward shift in
the number mode diameter can also be observed in
the smoldering samples, although it is, at best, only
about 0.1 micrometer in magnitude. A plot of cumulative volume less than 3 micrometer particle diameter is shown in Figure 50.8 for the same four
samples. Here again a slight upward shift of about
0.05 micron in the median volume diameter is observed in the smoldering phase smoke samples.
Carbon Distribution
The apportionment of carbon among the major products of combustion is shown in Table 50.4 for the
10,"-,--------------------,
FLAMING
SMOLDERING
i
i
10
10
10-q_~~TITm--r. .TITm--r. .TITm--r"TIT~
10-2
PARTICLE DIA (micron)
Figure 50.7 Particle number distributions for the first two
(flaming) and final two (smoldering) smoke plume penetrations
with the instrumented aircraft.
418
Chapter 50
flaming and smoldering phases of combustion. Gaseous volatile organic carbon compounds were not
quantified during this experiment. We have used a
model of Ward et al. (1990) to predict emission factors
for methane and nonmethane hydrocarbons. This
model is derived from measurements of combustion
efficiency as well as methane and nonmethane hydrocarbon emissions from a number of prescribed burns
in the Pacific Northwest. During both flaming and
smoldering conditions, in excess of96% of the carbon
is released as C02 and CO. During flaming conditions
the volume ratio of CO to C0 2 is 3.1 x 10- 2 • This
ratio increases to 1.2 x 10- 1 during smoldering conditions. Volatile and nonvolatile carbon particulate
material accounts for 0.85% of the total released carbon in the flaming phase and nearly 2.4% during the
100
90
~
80
z 70
smoldering phase. The first three categoriesnamely, C02 , CO, and volatile organic particulate
carbon-account for in excess of 99% of the carbon
released in both flaming and smoldering phases.
Estimates of Total Pollutant Release
We calculated the total release of various gas and
particle species during both flaming and smoldering
periods by breaking down the total fire interval into
flaming and smoldering periods and applying phasespecific emission factors (Table 50.3, columns 2 and 3)
and measured fuel consumption for each phase.
These estimates are given in units of kg m _, in Table
50.5. The total from flaming and smoldering phases is
compared to estimates of total release using average
emission factors (Table 50.3, column 4) from all aircraft samples collected and total fuel consumption.
Using the average CO yield results in a 64% overestimate of the total CO emissions when compared to the
X
•
Table 50.4
Species contribution to total mass release
of carbon from flaming and smoldering phases
fi2
Q
"'"
"'!!I
g"
-'
w
~
5
!i()
Carbon contribution
(% by weight)
60
50
Species
Flaming
Smoldering
Carbon dioxide
95.8%
86.6%
40
30
20
FLAMING
SMOLDERING
10
0
to-t
too
PARTICLE DIA (mfcron)
Figure 50.8 Cumulative volume ( <2.5 micron diameter) distribution for the first two and final two smoke samples.
Carbon monoxide
3.0
10.1
Organic carbon
particulate
0.7
2.3
Elemental carbon
particulate
0.2
0.1
Methanea
0.2
0.5
Nonmethane
0.1
0.4
hydrocarbonsa
a. Estimated from earlier work by Ward (1991).
Table 50.5 Total mass release of significant pollutants
Mass release (kg m -z)
Species
Flaming
Smoldering
Total"
Averageb
Ratioc
co,
co
21.5
0.43
0.062
0.010
0.035
0.031
0.037
8.0
0.60
0.055
0.003
0.045
0.031
0.058
29.4
1.03
0.117
0.012
0.080
0.062
0.095
28.3
1.69
0.133
0.010
0.102
0.076
0.141
0.96
1.64
1.13
0.77
1.22
1.22
1.48
0
EC
CH,
TNMHC'
PM2.5
a.
b.
c.
d.
Total is the sum of mass release from the flaming and smoldering categories.
Average is determined by use of average emission factors as measured over the duration of the fire.
Ratio is calculated with "Average" as the numerator and "Total" as the denominator.
1NMHC refers to total nonmethane hydrocarbons.
419
Einfeld, Ward, and Hardy
sum of CO yields from flaming and smoldering periods. Similarly, the PM2.5 emissions determined by
an average emission factor are about 50% higher than
the sum of the phase-specific emission yields. The
agreement for other carbonaceous species such as
volatile and nonvolatile carbon is somewhat better
and is likely a result of the fact that the difference
between flaming and smoldering emission factors is
less pronounced for these species.
Conclusions
Measurements of smoke emissions were made of the
0.43 km2 Down Creek fire in northwestern Montana
using an instrumented aircraft and smoke sampling
system flown periodically through the smoke column.
Results from these tests demonstrated an expected
change in the characteristics of the emissions as the
fire progressed from flaming to smoldering conditions
over a period of about two hours. Combustion efficiencies ranged from 95% conversion of released fuel
carbon to C02 during the flaming phase to 85% conversion during the smoldering phase. The overall fuel
mass consumption was about 70%, with consumption
of 81% of the woody fuel and 62% of the duff layer.
The median particle diameter became larger for
smoke derived from smoldering combustion sampled
toward the latter stages of the fire. Although these
particle size changes are not likely to be significant in
terms of the mass release of particulate matter and its
atmospheric transport, they are accompanied by
appreciable changes in smoke optical properties. An
enhanced scattering deficiency is observed in smoldering samples and is consistent with predictions
based on Mie theory for an upward shift in the particle
median volume diameter. This observation becomes
important in the context of radiative transfer through
the smoke plume.
A comparison of measured emission factors from
flaming and smoldering periods indicates about a
threefold increase in CO, PM2.5, and volatile organic
particulate carbon (for particles <2.5 micrometers in
diameter) for emissions from smoldering combustion.
These emission factor increases are at least partially
offset by a reduced fuel consumption during the smoldering phase. Total fuel consumption for the Down
Creek prescribed fire was 12.4 kg m - 2 for the flaming
phase and 5.0 kg m - 2 for the smoldering phase. An
approximate threefold increase in emission factors is
then offset by about a 2.5-fold reduction in fuel consumption in the smoldering phase. Taken together,
the net result is an approximate 50% increase in the
L
total release of such pollutant species as PM2.5 and
CO during smoldering conditions as compared to
flaming periods.
Results suggest that knowledge of fuel consumption by phase of combustion (flaming vs. smoldering)
is important in making accurate estimates of the characteristics of smoke emissions from individual fires.
Contributing factors such as fuel type, fuel loading,
and meteorological history vary significantly by
region and should be taken into account when compiling estimates of fuel consumption rates during both
flaming and smoldering fire conditions.
Acknowledgments
We recognize the assistance of Janet Hall and her
documentation of the prescribed fire history. We
thank the Champion Paper Company for their generosity in allowing this particular fire to be used for
detailed study. The assistance of the Libby Ranger
District of the U.S. Forest Service in site preparation
is also noted. We also recognize the Defense Nuclear
Agency, Global Effects Program, for their financial
support of this work.